MAG4’s New Database of HMI Active-Region Vector Magnetograms: Sample Size and Initial Results for Major-Flare Forecasting

We have developed a large-database method of forecasting an active region’s (AR’s) chance of producing a major flare (GOES M- or X-class) and its chance of producing a major CME (speed > 800 km/s) in the coming few days from a free-energy proxy – a proxy of the AR’s free magnetic energy – measured from a magnetogram of the AR. We have named this forecasting tool MAG4 (for Magnetogram Forecast). In its present near-real-time operation mode, MAG4 forecasts each on-disk AR’s rates of production of major flares and major CMEs in the coming few days, based on the observed major-flare and major-CME histories of 1,300 ARs observed within 30 degrees of disk center in MDI line-of-sight magnetograms. From the passages of these ARs across the 30-degree central disk, the presently-used MAG4 MDI database has the value of a free-energy proxy measured from 40,000 MDI magnetograms of these 1,300 ARs. We are now compiling a similar database of the about the same size for MAG4, but for HMI vector magnetograms that are of ARs observed within 45 degrees of disk center and that have been deprojected to disk center. This new MAG4 HMI database now has a wide variety of AR parameters measured from each of 40,000 deprojected HMI vector magnetograms of 900 ARs within 45 degrees of disk center (15 magnetograms of each AR per day during its passage across the 45-degree central disk). We present the MAG4 major-flare forecasting curves obtained from this new database for a few alternative free-energy proxies measured from either the vertical-field component or the horizontal-field component of the deprojected AR vector magnetograms. (The magnetogram’s horizontal-field component more directly reflects the AR’s free magnetic energy than does the magnetogram’s vertical-field component.) By using our statistical method of measuring, via a skill score, whether the forecasting performance of one AR magnetogram parameter is significantly better than that of another, we show which free-energy proxy is the best major-flare predictor that we have found so far.